Energy and Buildings 36 (2004) 987–993

www.elsevier.com/locate/enbuild

Assessment of productivity loss in air-conditioned

buildings using PMV index
R. Kosonena,*, F. Tanb,1
a
Halton OY, Haltonini 1-3, Kausala 47400, Finland
b
CapitaLand Commercial Limited, 39 Robinson Road, Singapore 068911

Abstract

This theoretical study reports on the assessment of productivity loss in air-conditioned office buildings using the PMVapproach and makes
use of Wyon’s reviews [D.P. Wyon, P.O. Fanger, B.W. Olesen, C.J.K. Pedersen, The mental performance of subjects clothed for comfort at two
different air temperatures, Ergonomics 18 (1975) 358–374; D.P. Wyon, Individual microclimate control: required range, probable benefits and
current feasibility, in: Proceedings of Indoor Air ’96, Institute of Public Health, Tokyo, 1996; D.P. Wyon, Indoor environmental effects on
productivity. IAQ 96 Paths to better building environments/Keynote address. Y. Kevin. Atlanta, ASHRAE, pp. 5–15] as the basis to compare
and to relate how the productivity loss could be minimised through improved thermal comfort design criteria. The finding shows that task-
related performance is significant correlated with the human perception of thermal environment that in turn is dependent on temperatures.
Different combinations of thermal criteria (air velocity, clo, metabolic, etc.) can lead to similar PMV value and the PMV equation is useful to
predict productivity loss that is due to the rate of change in thermal conditions. The study also highlights the issues that remain to be resolved
in future research.
# 2004 Elsevier B.V. All rights reserved.

Keywords: Productivity; PMV index; Air-conditioned offices

1. Introduction operating a building in developed countries; (2) there is a

Air-conditioned office buildings aim to provide a ther-
mally acceptable environment for human comfort and work
that would in turn enable better work productivity and less
thermal dissatisfaction. The initial investment cost of an air-
conditioning system is the usual first criterion in its system
selection. This is a limited approach because it can be costly
in the building economic life cycle if the air-conditioning
operation and maintenance cost and the impact on office
workers’ productivity are not duly considered.
A literature review has identified the following findings:
(1) the salaries of office workers are many times the cost of
* Corresponding author. Present address: Haltoninitie 1-3,
Kausala 47400, Finland. Tel.: +358 5 740 211; fax: +358 5 740 2500.
E-mail address: risto.kosonen@halton.com (R. Kosonen),
freddietan@capitaland.com.sg (F. Tan).
1
Present address: 39 Robinson Road, Robinson Point,
Singapore 068911. Tel.: +65 6239 6884; fax: +65 6876 6495.
potential monetary gain due to improved workers’ produc-
tivity if there is a shift in focus towards the provision of
better indoor climate conditions; (3) improving working
efficiency in thermal indoor environment is deemed to be
the most important environmental factor in office produc-
tivity study. Woods [4] has determined that the salaries of
workers in the US office buildings have exceeded the cost of
building energy, maintenance, annualized construction and
rental by a factor of 100. Skåret [5] has published a similar
study in Norway and estimated that increased productivity
due to an improved indoor climate is at least 10 to 100 times
greater than the operational and maintenance costs.
More recent US studies such as [6,7] have estimated
that the yearly potential gain of productivity increase due
to the reduction of respiratory infection would equal US$ 6–
14 billion while a reduction of sick building syndromes
(SBS) could yield around US$ 15–38 billion. Most sig-
nificantly, improved working efficiency could yield US$

0378-7788/$ – see front matter # 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.enbuild.2004.06.021
988 R. Kosonen, F. Tan / Energy and Buildings 36 (2004) 987–993
20–200 billion. Similar findings have been published else-
where in developed countries, such as a reported Finnish
study by Seppanen [8] that estimated a loss of 2.7 billion
Euros due to poor indoor climate conditions in Finland. In
Asia, a recent study by Gan and Tan [9] has emphasized the
negative consequences of a poor indoor thermal environ-
ment and stated that the use of life cycle costing approach is
necessary to mitigate the monetary implication of thermal
comfort and productivity in tropical air-conditioned build-
ings. The productivity loss issue is therefore prevalent in
poorly designed air-conditioned enclosures regardless of
external climate variation and thus a primary aim in this
study is to design quality indoor thermal conditions that are
suitable for the envisaged tasks in office spaces, thus pro-
moting better productivity.
This theoretical study reports on the assessment of pro-
ductivity loss in air-conditioned office buildings using the
predicted mean vote (PMV) approach and makes use of
Wyon’s reviews [1–3] as the basis to compare and to relate
how the productivity loss could be minimised through
improved thermal comfort design criteria. This ‘retrospec-
tive’ interpretation, by means of PMV parametric analysis,
also helps to re-examine the nature of productivity loss
reported in earlier studies.
2. Thermal environment
Thermal comfort is a fluid concept and is derived from
the actual requirement of people. A human being’s thermal
sensation is related to the thermal balance of his body as a
whole. Using the first principle of adaptive thermal comfort
approach, the subject will secure his own comfort through
behavioural adjustment. The balance is influenced by his
metabolic heat production, physical activity, clothing and
the four environmental parameters: air temperature, mean
radiant temperature, air velocity and air humidity. At the
outset, occupants in an air-conditioned building could nor-
mally adapt to the indoor thermal environment through
physiological means to attain their own comfort. When
these factors are known, the thermal sensation for the body
as a whole can be predicted by the predicted mean vote
calculation. There could possibly be different combinations
of thermal criteria at thermal neutrality (PMV = 0) that could
give rise to the same work performance (productivity). One
such example is reported by Wyon et al. [1] where the
conditions of 0.6 clo value at 23.2 8C air temperature and
1.15 clo at 18.7 8C air temperature are cited to lead to similar
productivity level.
Hence, the PMV concept [10] is relevant as a tool in the
assessment of productivity loss for the rate of change in
thermal comfort criteria. This could provide a better insight
into the thermo-physics of the heat exchange between an
occupant and his environment; and the impact on pro-
ductivity in an indoor air-conditioned office environment.
Several other studies have shown that very moderate heat
stress can negatively affect mental performance [3]. In the
case of a warm thermal environment, the blood vessels
would normally vasodilate, increasing the blood flow
through the skin and at higher thermal load one begins to
perspire. In the absence of conscious effort, the human body
might tend to adapt through lowering of internal heat
production and this reduces or even avoids perspiration.
This subconscious behavioural adjustment could mean a
lower arousal and results in a slower work rate. This is the
responsiveness of human thermoregulatory system towards
external variation in climate, work condition and perceived
sensations.
3. Productivity measure
It should be noted that optimal performance (productiv-
ity) does not necessarily occur under optimal thermal com-
fort conditions and this is supported by evidence in an early
study [11] where subjects were performing mental work at
air temperatures in the range 20–30 8C. A more recent report
by Wyon [3] has summarized numerous thermal effects on
performance in laboratory environments. Productivity loss is
presented as a function of room temperature in different
tasks. The main findings are two-fold: (1) for thinking tasks,
the rate of working was reduced by 30% at 27 8C when the
average neutral temperature was 21 8C and remained con-
stant at 70% for higher temperatures above 27 8C and (2) for
typing tasks, performance decreased to 70% over 4 8C and
remained at 70% at higher temperatures. Thus, productivity
loss is strongly dependent on the nature of the performed
task. The approach makes use of two main tasks: typing and
thinking. The best approach is to fit the general pool of job
descriptions by using the time weighting factors for the
tasks.
These past findings provides the impetus to model the
simulated subjective responses of occupants in buildings,
using the PMV parametric analysis, with respect to the rate
of change in thermal conditions and the impact of the
changing thermal conditions on both the thinking and typing
tasks where the thermal neutral balance situations are fully
documented. Wyon’s results [1–3] are therefore used and re-
constructed in this study to create a generic productivity loss
model using the PMV index as a ‘navigating’ search and
assessment tool. The neutral temperature concept is used in
this assessment of productivity loss in air-conditioned office
buildings.
4. Base case comparison to new simulated situations
Table 1 reports the results of the parametric analysis that
is carried out at a room temperature of 21 8C and relative
humidity of 50%, in a condition that is somewhat akin to
Wyon’s review [3]. The Cases 1–4 in Table 1 are examples of
the possible thermal neutral conditions in the test. In the
 R. Kosonen, F. Tan / Energy and Buildings 36 (2004) 987–993 989

Table 1
Comparison of PMV values at different temperature differences from the neutral conditions in the selected cases
Estimated conditions for Wyon’s review (base case) Other simulated cases
1 2 3 4
DTrad 0 8C 0 8C 0 8C 2 8C 2 8C
v 0.15 (m/s) 0.15 (m/s) 0.25 (m/s) 0.15 (m/s) 0.15 (m/s)
clo 1.16 1.2 1.25 1.23 1.03
M 1.2 1.1 1.2 1.2 1.2
Temperature difference Reconstructed PMV results Other simulated
using a starting base cases – PMV results
temperature of 21 8C

0 0 0 0 0 0
+1 0.21 0.21 0.21 0.19 0.23
+2 0.42 0.43 0.42 0.39 0.46
+3 0.63 0.65 0.63 0.59 0.68
+4 0.84 0.87 0.85 0.79 0.91
+5 1.06 1.09 1.07 0.99 1.14
+6 1.28 1.32 1.28 1.19 1.38
DTrad: temperature difference (K) between room temperature and mean radiant temperature; v: air velocity (m/s); clo: clothing value (clo); M: metabolic rate
(W/m2); PMV: predicted mean votes.

selected situations (Cases 1–4), the basic assumptions are as
follows:

air velocity is 0.15 m/s (except case 2: 0.15 m/s);

radiant temperature is assumed to be equal to the room
temperature (except Case 3: DTrad = 2 8C and Case 4:
DTrad = +2 8C);

metabolic rate is 1.2 MET (except Case 2: 1.1 MET);

relative humidity is 50%;

room (air) temperature is 21 8C.
Based on these values in the selected cases, the requested
clo values are calculated at the base temperature level of
21 8C to reach thermal neutrality.
In the base case, the clo value has been adjusted to 1.16 to
reach the thermal neutral conditions (PMV = 0). At the same
way in Cases 1–4, the requested clo values are calculated to
reach thermal neutrality.
The effect of the temperature rise from the neutral
conditions for PMV index is slightly different with the
different combinations of thermal criteria. The biggest
difference occurs when there is deviation between the
radiant temperature and the room temperature. Cases 3
and 4 in Table 1 demonstrate a change in the radiant
temperature to attain the thermal neutrality at a room air
temperature 21 8C. Anyhow, the PMV values at different
room air temperatures of Cases 1 and 2 differ very little from
the base case which is estimated to describe the conditions in
the reviewed study.
This finding has demonstrated the followings:

different combinations of thermal criteria (air velocity,
clo, metabolic, etc.) lead to similar PMV values at dif-
ferent temperature levels when the radiant temperature is
assumed to equal to the room temperature;

the reconstructed conditions are reasonable accurate for
use to predict the PMV-productivity loss analysis and
generalize the previous temperature correlation to cover
PMV and productivity interdependence.
5. PMV prediction of thinking and typing tasks in office
The above base case comparison has shown that the PMV
parametric analysis approach is possible to create the causal

Table 2
PMV prediction of thinking and typing tasks at different temperatures
Temperature difference from PMV prediction Predicted percentage Wyon’s productivity Wyon’s productivity
base temperature of 21 8C dissatisfied, PPD (%) loss in thinking (%) loss in typing (%)
1 0.21 6.3 0.0 0.0
0 0 5.0 1.9 4.9
+1 0.21 6.0 5.4 13.5
+2 0.42 9.0 10.1 23.2
+3 0.63 14.0 15.3 30.0
+4 0.84 21.0 20.6 32.8
+5 1.06 29.0 25.7 33.4
+6 1.28 38.0 30.0 33.5
990 R. Kosonen, F. Tan / Energy and Buildings 36 (2004) 987–993
relationship between the PMV index and results of the
productivity loss for thinking and typing office tasks that
are reported by Wyon [3] (Table 2).
Using the predicted PMV results, the peak level of the
productivity (assume = 100%) occurs when PMV value is
0.21 at a temperature of 20 8C. The PPD in this case has a
low value of 6.3%. This hypothetical benchmark is aimed to
offer a better insight of thermal responses at the correspond-
ing higher temperatures. The results show that, as the
temperature increases above the neutral temperature of
21 8C for the envisaged productivity losses in thinking
and typing, the corresponding PMV values move from 0
(neutral sensation) to 1.28 (warm sensation). At a tempera-
ture of 24 8C, the productivity loss in typing task exceeds
30% with a predicted PMV value of 0.63 (slightly warm
sensation). This is significantly different compared to a
thinking task where 30% or more productivity loss only
starts to occur at a higher temperature of 27 8C.
This finding implies the followings:

task-related performance is significant correlated with the
human perception of thermal environment that in turn is
dependent on PMV and PPD values;

the value of the productivity loss in thinking is quite close
to PPD value and in typing about two times higher than
PPD value when PMV is less than 0.65;

normally accepted PMV value of +0.5 leads about 12%
productivity loss in thinking and 26% loss in typing.
Using the mathematical expression of Productivity Loss,
y = c0(PMV) + c1(PMV)2 + c2(PMV)3 + c3(PMV)4 +    +
cn(PMV)n. The following two Eqs. (1) and (2) are derived for
typing and thinking tasks (Fig. 1).
For typing task:
y ¼ 60:543x6 þ 198:41x5  183:75x4  8:1178x3
þ 50:24x2 þ 32:123x þ 4:8988
Fig. 1. Curve fitting of thinking and typing tasks.
For thinking task:
y ¼ 1:5928x5  1:5526x4  10:401x3 þ 19:226x2
þ 13:389x þ 1:8763 (2)
Using the curve fitting equations for thinking and typing
tasks, it is possible to calculate the productivity loss as a
function of PMV, and as a function of PPD using the typing
and thinking tasks in Eqs. (1) and (2), respectively as the
boundary constraints. The results are shown in Figs. 2 and 3
below.
The envelope of curves shown in Figs. 2 and 3 arising out
of the mathematical expressions derived in Eqs. (1) and (2) is
fairly accurate to predict the productivity loss vis-à-vis the
traditional PMV and PPD approaches.
The nature of the task is a key determinant for the rate of
productivity loss. The productivity of the worker decreases
at a sharper rate in typing tasks than thinking tasks. For
example, if the share of thinking and typing is 50:50 with a
PMV value of 0.5, the productivity loss is about 6% higher
than pure thinking task.
It is noted that an approximate linear correlation between
PPD and productivity loss (20%) occurs for thinking tasks.
Conversely, in typing tasks, the productivity loss (20%)
can be almost twice as much as the PPD. (See Fig. 3
comparison for typing task: PPD =15% and productivity
loss = 30% and thinking task: PPD = 15% and productivity
loss = 15%.)
6. Neutral temperatures and its impact on productivity
The acceptable PMV range is usually between 0.5
(slightly cool sensation) and +0.5 (slightly warm sensation).
In tandem with the acceptable PMV range, the respective
neutral temperatures of 21, 23 and 25 8C would produce a
temperature range of +2.5, +2.0 and +1.5 8C, respectively.
(1) This condition is only valid when all other parameters are
constant.
In practice, there is a possibility to adjust clothing
insulation at lower temperatures and in turn this helps to
increase the neutral temperature range. At higher tem-
peratures, this is a limited possibility because the clothing
value has to conform to the social milieu. For example,
the necessary clo value would be approximately 0.84 at
23 8C and 0.53 at 25 8C. Adjustment methods, such as
a move relaxed clothing norm that result in lower clo
value, approximate to 0.50 clo, could be used to achieve a
higher neutral temperature, but it is not possible in all
cases.
Figs. 4 and 5 show that the decline of productivity at the
different neutral temperatures for both typing and thinking
tasks. The envisaged productivity for typing tasks will
decline to 70% when the temperature difference between
room temperature and neutral temperature is 3 8C. For
thinking tasks, the tolerance threshold is higher than that
 R. Kosonen, F. Tan / Energy and Buildings 36 (2004) 987–993 991

Fig. 2. Productivity loss as a function of PMV for different combination of tasks.

Fig. 3. Productivity loss as a function of PPD for different combination of tasks.

of typing tasks. The temperature difference for a 30% lost of
productivity is 4–6 8C depending on the neutral temperature.
7. Productivity impact of relative humidity
Relative humidity may also affect productivity. A max-
imum relative humidity between 50 and 65% is normally
provided at the design stage. A higher relative humidity
affects the perceived indoor air quality and thus affects
thermal comfort. Fig. 6 psychrometric chart is used to
illustrate the changes that would occur from an air tem-
perature of 25 8C and relative humidity of 50%. The
optimal productivity is around 24 8C at a predicted
PMV value of 0.21 (Fig. 6). The first line illustrates
that the productivity loss is under 5% and the second line
shows a 15% productivity loss. In Fig. 6 shows ASHREA
comfort range.

Fig. 4. Neutral temperatures and their impact on productivity for a thinking Fig. 5. Neutral temperatures and their impact on productivity for a typing
task. task.
992 R. Kosonen, F. Tan / Energy and Buildings 36 (2004) 987–993

Fig. 6. Psychrometric chart illustrating optimal comfort and productivity loss lines of 5 and 15% compared with the ASHREA comfort zone.

The productivity loss of 5% is equivalent with 0–1.5 8C 8. Cost implication of productivity

temperature rise depending on the relative humidity. A 15%
productivity loss means a 1.0–2.5 8C temperature rise. It
should be noted that Fig. 6 illustrates slightly stronger
correlation between the productivity loss and the relative
humidity than ASHREA describes. Based on the used
model, the room temperature should be about 0.5 8C lower
at high humidity. This means that the ASHREA limit: room
temperature 26 and 17 8C dew point should be instead 25.5
and 17 8C.
The effect of the humidity is higher when the neutral
temperature is higher. Table 3 shows the effect of the
humidity when the thermal neutral condition is 25 8C and
50%. At the same temperature level of 25 8C, the produc-
tivity loss increases about 2% when the humidity increases
from 50 to 75%. At the 27 8C level, a decrease of 5% in
productivity occurs.
The initial investment cost of an air-conditioning system is
the usual first criterion in system selection. This is a limited
approach because it can be costly in the building economic life
cycle if the air-conditioning operation and maintenance cost
and the impact on office workers’ productivity are not duly
considered: (1) The salaries of office workers are many times
the cost of operating a building in developed countries; (2)
there is a potential monetary gain due to improved workers’
productivity if there is a shift in focus towards the provision of
better indoor climate conditions; (3) improving working
efficiency in thermal indoor environment is deemed to be
the most important environmental factor in office productiv-
ity. Fig. 7 shows the influence of the temperature difference

Table 3
Effect of relative humidity on productivity (25 8C and 50% is the thermal
neutral condition and at 24 8C and 50% the productivity is maximum)
Relative humidity Productivity Relative humidity Productivity
at temperature loss (%) at temperature loss (%)
25 8C (%) 27 8C (%)
35 0.7 35 12.1
50 1.9 50 15.0
55 2.4 55 16.1
65 3.4 65 18.0
Fig. 7. Cost implication of productivity loss at different neutral tempera-
75 4.8 75 20.0
tures.
 R. Kosonen, F. Tan / Energy and Buildings 36 (2004) 987–993
from the neutral conditions for different annual salary costs
and different neutral conditions. Two different salaries are
used in the figure.
Fig. 7 shows that a small difference of 2 8C between room
temperature and neutral temperature could result in more
than 10% productivity loss. In monetary terms, this is about
US$ 3000 with US$ 25,000 salary and about US$ 10,000
with US$ 75,000 salary for a drift of 2.0 8C. If we consider a
period of 10 years, the productivity loss is equivalent to one
year’s salary. The life-cycle cost of HVAC system for a
period of 10 years is about US$ 200/m2. Using the spatial
planning and use of 10 m2 per person, a comparison between
the productivity loss and HVAC life-cycle costs indicates
that the productivity loss of a 2 8C temperature difference
from the neutral conditions could easily be 2.5–5 times
higher than the cost of the system over the life cycle. At
this design criteria, the envisaged productivity loss is
approximately US$ 2000 per person (10 m2 per person 
US$ 200/m2) for a 10 m2 of space in office buildings each
year.
9. Inferences from current study

Previous laboratory studies [2,3] are relevant to a neutral
temperature range between 20 and 27 8C. The productiv-
ity loss at a temperature beyond 27 8C has been assumed
to be constant.

The level of the neutral temperature is an important factor
to indicate the rate of the productivity change (increase or
decrease) when the room temperature would increase.

The acceptable PMV range is usually between 0.5 and
+0.5. However the PMV value of +0.5 leads significant
productivity loss.

In future experiments or tests, PMV values should be fully
documented.

The current PMV estimation is based on the causal
relationship between temperatures and productivity;
and is limited to experiments that have been carried
out in temperate environments. The true implication for
hot-and-humid tropical climates in air-conditioned build-
ings remains relatively unknown.

Though relative humidity is a less important factor for
thermal comfort-productivity loss in temperate climates,
this is a crucial factor to understand the actual thermal
comfort requirements in the hot-and-humid climates that
in turn impact the level of productivity in offices.

ASHREA comfort range slightly underestimates the influ-
ence of the humidity at high humidity level.
10. Conclusion
As a general rule, the productivity loss when thinking is
less severe than the productivity loss when typing in a
warmer environment. Task-related performance is signifi-
993
cant correlated with the human perception of thermal
environment that in turn is dependent on temperatures.
Different combinations of thermal criteria (air velocity,
clo, metabolic, etc.) can lead to similar PMV values. The
PMV equation is useful to predict productivity loss that is
due to the rate of change in thermal conditions. The refer-
ence temperature (neutral temperature) setpoints for typing
and thinking tasks may be helpful to assess productivity loss
in different climates but nonetheless require further verifica-
tion. This is especially true in the hot-and-humid climates,
particularly those in the Southeast Asia region, where there
is a pressing need to evaluate thermal comfort standards due
to an escalating demand for air-conditioning.
Acknowledgements
The co-authors take this opportunity to thank their
respective organizations, Halton Group (Finland) and PRE-
MAS International Limited (Singapore) for their support in
this work. The respective co-authors also wish to acknowl-
edge the past support coming from the Finnish Technology
Agency (TEKES) and Singapore Agency for Science, Tech-
nology and Research (A*Agency), respectively. The authors
thank Dr. David Wyon for many insightful comments on a
draft of this paper.
References
[1] D.P. Wyon, P.O. Fanger, B.W. Olesen, C.J.K. Pedersen, The mental
performance of subjects clothed for comfort at two different air
temperatures, Ergonomics 18 (1975) 358–374.
[2] D.P. Wyon, Individual microclimate control: required range, probable
benefits and current feasibility, in: Proceedings of Indoor Air ’96,
Institute of Public Health, Tokyo, 1996.
[3] D.P. Wyon, Indoor environmental effects on productivity. IAQ 96
Paths to better building environments/Keynote address, Y. Kevin.
Atlanta, ASHRAE, 1996, pp. 5–15.
[4] J.E. Wood, Cost avoidance and productivity in owning and operating
buildings, Occupational Medicine 4 (4) (1989) 753–770.
[5] J.E. Skåret, ‘‘Indoor environment and economics’’; Project no. N
6405, The Norwegian Institute of Building Research (NBI-Byggforsk)
(in Norwegian), Oslo, February 1992.
[6] W. Fisk, A. Rosenfeld, Estimates of improved productivity and health
from better indoor environments, International Journal of Indoor Air
Quality and Climate 7 (1997) 158–172.
[7] W. Fisk, A. Rosenfeld, Potential nationwide improvements in pro-
ductivity and health from better indoor environments, in: Proceedings
of the 1998 Summer Study on Energy Efficiency in Buildings,
American Council for an Energy-Efficient Economy, 1998.
[8] O. Seppänen, Estimated cost of indoor climate in Finnish buildings, in:
Proceedings of the Eighth International Conference on Indoor Air
Quality and Climate, Edinburgh, Scotland, 1999.
[9] C.E. Gan, F. Tan, Cost evaluation of thermal comfort in tropical office
buildings, in: Proceedings of the CIB International Conference on
performance-based buildings, Hong Kong, 5–10 May, 2002.
[10] P.O. Fanger, Thermal comfort: analysis and application in environ-
mental engineering Danish Technical Press, 1970.
[11] R.D. Pepler, R.E. Werner, Temperature and learning: an experimental
study, ASHREA Transactions 42 (1968) 211–219.